CHARACTERIZATION OF SEDIMENT PARTICLE SIZE AND SHAPE USING MULTIVARIATE DATA ANALYSIS METHODS

The size and shapes of clastic particles have historically been used to help determine the geological provenance and mechanical behavior of sediments and the transport processes to which the particles have been exposed. Optical microscopy and/or scanning electron microscopy (SEM) have often been used to observe the outlines of particles, which are then qualitatively categorized based on the experience of the observer using terms such as rounded, subrounded, subangular, angular, flaky, etc. Such measurements are subject to the experience and judgement of the observer and are time consuming and tedious.

The goal of this work has been to use optical microscopy to automatically characterize the particles in Mars analog sediments using image analysis and various methods of measuring the shape of individual particles. Particle shape can be described using many different parameters including simple geometrical measurements and the components of the Fourier transform of the particle outline. Once the shape parameters of many particles have been calculated from their images, various methods of analyzing this multivariate data such as principal component analysis (PCA) and partial least squares (PLS) fitting are applied.

The application of PCA and PLS to common geometrical shape descriptors discussed extensively in the literature (e.g. perimeter, area, feret diameters, sphericity, etc.) has demonstrated that only a few descriptors are unique enough to provide adequate differentiation between sediment samples. The high frequency components of the Fourier transform of a particle outline can be used to provide additional unique shape parameters. The application of multivariate methods to this more complete set of shape descriptors provides more complete separation of the particle ensembles that comprise sediments from different locations.

We have applied this method to samples of terrestrial sediments that can be considered analogs for Martian aeolian and fluvial sediments in order to help define the requirements for an automated microscope for characterizing Martian sediments in-situ. Current “microscopes” being readied for deployment on Mars have a maximum magnification of approximately 12X. Our results demonstrate the need for in-situ microscopes of considerably higher magnification.